Relations Between Cognitive Abilities and Measures of Executive Functioning Timothy A. Salthouse

advertisement
Neuropsychology
2005, Vol. 19, No. 4, 532–545
Copyright 2005 by the American Psychological Association
0894-4105/05/$12.00 DOI: 10.1037/0894-4105.19.4.532
Relations Between Cognitive Abilities and Measures of
Executive Functioning
Timothy A. Salthouse
University of Virginia
Although frequently mentioned in contemporary neuropsychology, the term executive functioning has
been a source of considerable confusion. One way in which the meaning of a variable can be investigated
involves examining its pattern of relations with established cognitive abilities. This method was applied
to a variety of variables hypothesized to assess executive functioning in 2 data sets, 1 consisting of 328
adults between 18 and 93 years of age and a 2nd composite data set based on nearly 7,000 healthy adults
between 18 and 95 years of age. Most of the hypothesized executive functioning variables were strongly
related to reasoning and perceptual speed abilities, and very few had any unique relations with age after
taking into consideration the relations of age through the cognitive abilities. These results raise questions
about the extent to which neuropsychological tests of executive functioning measure a distinct dimension
of variation in normal adults.
Keywords: executive functioning, aging, cognitive ability, reasoning, perceptual speed
itoring, sequencing, and other supervisory processes which permit
the individual to impose organization and structure upon his or her
environment” (Foster, Black, Buck, & Bronskill, 1997, p. 117).
“The executive functions consist of those capacities that enable a
person to engage successfully in independent, purposive, selfserving behavior” (Lezak, 1995, p. 42). “Executive functions include the following abilities: 1. Formulating goals with regard for
long-term consequences. 2. Generating multiple response alternatives. 3. Choosing and initiating goal-directed behaviors. 4. Selfmonitoring the adequacy and correctness of the behavior. 5. Correcting and modifying behaviors when conditions change. 6. Persisting in the face of distraction” (Malloy, Cohen & Jenkins, 1998,
p. 574).
Perry and Hodges (1999) “refer to executive functions as those
higher-order cognitive capabilities that are called upon in order to
formulate new plans of action and to select, schedule, and monitor
appropriate sequences of action” (p. 389). “The ‘executive functions’ broadly encompass a set of cognitive skills that are responsible for the planning, initiation, sequencing, and monitoring of
complex goal-directed behavior” (Royall et al., 2002, p. 378).
According to Tranel, Anderson, and Benton (1994), “the term
executive functions . . . [is] . . . used to refer to higher-order
cognitive capacities, for example, judgment, decision-making,
planning, and social conduct” (p. 126). “Executive functioning
involves problem solving abilities such as abstraction, planning,
strategic thinking, behavioral initiation and termination, and selfmonitoring” (Troyer, Graves, & Cullum, 1994, p. 45). “Executive
functions are often referred to as the most complex of human
behaviors being primarily concerned with planning and organization of purposeful behavior” (Tuokko & Hadjistavropoulos, 1998,
p. 143).
Many of these descriptions are quite broad, and several seem to
refer to rather general aspects of thinking. It is thus not surprising
that a number of researchers have speculated that a close relation
might exist between executive functioning and the concept of fluid
intelligence (e.g., Denchla, 1995; Duncan, 1994; Duncan, Emslie,
Williams, Johnson, & Freer, 1996; Duncan, Johnson, Swales, &
Executive functioning is hypothesized to involve the control and
coordination of cognitive operations and has become an important
concept in contemporary neuropsychology. However, there is still
little consensus on what executive functioning actually means.
Consider the following sample of variables that have been used to
assess executive functioning in recent articles: Verbal Fluency
(Bastin & van der Linden, 2003; Mungas, Reed, & Kramer, 2003),
Trail Making (e.g., Barnes, Yaffe, Satariano, & Tager, 2003;
Bigler et al., 2003), Stroop Color Word (e.g., Barnes et al., 2003;
Bastin & van der Linden, 2003), and Digit Symbol or Symbol
Digit (e.g., Barnes et al., 2003; Bigler et al., 2003; Verghese et al.,
2003). Moreover, two recent neuropsychological test batteries
designed to assess executive functioning only have tests of verbal
fluency and sorting or categorization in common, with one battery
also including tests of proverb interpretation, inferring word meaning from context, and 20 questions (Delis, Kaplan, & Kramer,
2001) and the other battery also including tests of judgment and
maze solution (Stern & White, 2003). The diversity of variables
used to assess executive functioning suggests that there is not yet
much agreement about the nature of this construct and even less
about how it is best assessed.
The uncertainty about the nature of executive functioning is also
apparent in the following characterizations from recent articles and
books. “Executive functions cover a variety of skills that allow one
to organize behavior in a purposeful, coordinated manner, and to
reflect on or analyze the success of the strategies employed”
(Banich, 2004, p. 391). “Executive functions are those involved in
complex cognitions, such as solving novel problems, modifying
behavior in the light of new information, generating strategies or
sequencing complex actions” (Elliott, 2003, p. 50). “Executive
functions include processes such as goal selection, planning, mon-
This research was supported by National Institute on Aging Grant AG
RO1-19627 to Timothy A. Salthouse.
Correspondence concerning this article should be addressed to Timothy
A. Salthouse, Department of Psychology, University of Virginia, Charlottesville, VA 22904. E-mail: [email protected]
532
533
EXECUTIVE FUNCTIONING
Freer, 1997; Mittenberg, Seidenberg, O’Leary, & DiGiulio, 1989;
Phillips, 1997; Rabbitt, 1997; Salthouse, Atkinson, & Berish,
2003).
In light of the apparent uncertainty regarding the nature of
executive functioning, it is worth considering how the validity of
potential executive functioning measures might be evaluated. At
the current time, the primary basis for deciding that a test assesses
executive functioning is a decision that the test has face validity,
but judgments of this type are inherently subjective. A more
objective type of validity is predictive validity, but that is not
feasible with executive functioning because there is no external
criterion against which to compare measures of executive functioning (e.g., Burgess, 1997; Royall et al., 2002; Tranel et al.,
1994). Many purported executive functions are impaired in certain
patients with frontal lobe damage, and hence these functions are
sometimes linked to activity of the frontal lobes. Results from
functional neuroimaging studies might also be used to provide
evidence that tasks hypothesized to reflect executive functioning
represent something distinct from what is involved in other types
of tasks (e.g., Collette & van der Linden, 2002). However, because
the correspondence between brain region and function is not one to
one, tests cannot be validated as methods of assessing a particular
theoretical construct simply by finding that they are associated
with damage in a particular region of the brain (e.g., Duncan et al.,
1996; Foster et al., 1997; Perry & Hodges, 1999; Reitan & Wolfson, 1994; Robbins et al., 1997; Tranel et al., 1994).
One well-accepted method of investigating the meaning of a
variable from the psychometric tradition is based on the examination of construct validity. That is, a variable can be inferred to be
similar to other variables to which it is moderately related (indicating that it has convergent validity) and to be dissimilar to
variables to which it is only weakly related (indicating that it has
discriminant validity). Unfortunately, many of the studies in which
multiple measures of executive functioning were examined have
found weak or mixed results (e.g., see citations in Royall et al.,
2002, and in Salthouse et al., 2003). Results from a recent project
by Salthouse et al. (2003) are fairly typical. These researchers
investigated the construct validity of several commonly used measures of executive functioning including the Wisconsin Card Sorting Test (Heaton, Chelune, Talley, Kay, & Curtis, 1993), the
Tower of Hanoi (Salthouse et al., 2003), a variant of the Trail
Making Test, and tests of verbal and figural fluency. Although the
variables had moderate correlations with one another, this type of
convergent validity evidence was not considered very compelling
because most reliably assessed cognitive variables are positively
correlated with one another. Because the variance common to the
purported executive functioning variables was very strongly related to other cognitive ability constructs and particularly to a fluid
intelligence construct, it was concluded that, at least when assessed
with these variables, the construct of executive functioning had
little discriminant validity.
An alternative method of investigating the meaning of an individual variable involves examining relations of the variable with
established cognitive abilities and with an individual-difference
characteristic such as age. Consider the analytical model portrayed
in Figure 1. The boxes in this type of structural equation model
correspond to observed variables, and the circles represent latent
variables, or theoretical constructs, that are assumed to be responsible for at least some of the variation in the observed variables.
Unidirectional arrows portray an influence from one variable to
another, and bidirectional arrows represent a correlation without
any assumption about the causal direction between the variables.
When the coefficients for the relations from the cognitive ability
latent constructs to the target variable are standardized, they can be
interpreted as indicating the relative contribution of each ability to
the variance in the variable. That is, these coefficients provide an
objective estimate of the degree to which the variables are pure
with respect to particular cognitive abilities. The direct relation
from age to the target variable in this type of model is also of
interest because it indicates the unique contribution of age on the
variable, independent of any indirect age-related influences operating through the abilities.
The model portrayed in Figure 1 can therefore be used to
investigate the meaning of variables postulated to represent executive functioning by determining the degree to which the variables
are related to other cognitive abilities and the magnitude of any
unique relations to age. The rationale is that if a variable represents
a construct, such as executive functioning, that is distinct from the
other cognitive ability constructs included in the analysis, then the
coefficients from the ability constructs to the variable should be
relatively small. Furthermore, if the hypothesized construct is
assumed to have significant relations with age and if at least some
of the age-related influences on that construct are statistically
independent of the age-related influences on the other cognitive
ability constructs included in the analysis, then the coefficient for
the direct relation from age to the variable should be significantly
different from zero.
The primary goal of the current project was to use the analytical
model represented in Figure 1 to investigate relations of cognitive
abilities to variables often assumed to reflect executive functioning. The model was applied to two separate data sets. The first data
set was based on a new study involving 328 adults who performed
all of the 16 reference cognitive tasks as well as several tasks often
hypothesized to represent executive functioning, that is, the Wisconsin Card Sorting Test, verbal fluency tasks, and a variant of the
Trail Making Test. The second data set was based on data combined across multiple samples rather than on data from a single
study. This aggregate data set has several advantages over data
from a single study, including a much larger and more diverse
sample than is typically feasible in a single study and a broader
variety of target variables (cf. Salthouse, 2004).
Data Set 1
Summary characteristics of the sample for the first data set are
presented in Table 1. The participants were recruited from posted
flyers and newspaper advertisements, and there were 89 individuals between 18 and 39 years of age, 133 individuals between 40
and 59 years of age, and 106 individuals between 60 and 93 years
of age. Inspection of the table indicates that, on average, these
individuals were highly educated and in good health. The sample
was functioning at a relatively high cognitive level because the
scaled scores for the vocabulary and digit symbol variables
were 12.8 and 11.0, respectively. (Note that the scaled scores have
means of 10 and standard deviations of 3 in the nationally representative sample used to establish the test norms; cf. Wechsler,
1997a.)
Further examination of Table 1 reveals that all of the reference
variables for the cognitive abilities had acceptable levels of reliability and that with the exception of the vocabulary variables, they
534
SALTHOUSE
Figure 1. Schematic illustration of an analytical model used to investigate relations of cognitive abilities on a
target cognitive variable and the direct and indirect (mediated through the abilities) influences of age on the target
cognitive variable. Voc ⫽ vocabulary; WAIS ⫽ Wechsler Adult Intelligence Scale—Third Edition; WJ ⫽
Woodcock–Johnson Psycho-Educational Battery—Revised (Woodcock & Johnson, 1990); Reas ⫽ reasoning;
Space ⫽ spatial visualization; Mem ⫽ episodic memory; Ravens ⫽ Advanced Progressive Matrices: Set II
(Raven, 1962); Shipley ⫽ Shipley Institute of Living Scale—Revised (Zachary, 1986).
all had significant negative correlations with age. The structural
equation models were analyzed with the AMOS (Arbuckle, 2003)
computer program. Fit statistics for the model portrayed in Figure 1 with five correlated factors are presented in the bottom of
Table 1. Although the chi-square statistic provides a direct test of
differences between the predicted and observed variances and
covariances, it is usually significant with moderate sample sizes
and thus is not considered very diagnostic. Two more informative
statistics are the comparative fit index (CFI), which represents the
improvement relative to an alternative model, and the root-meansquare error of approximation (RMSEA), which represents the
deviation between the predicted and observed variances and covariances. Models with reasonable fit should have CFI values
greater than .9 and RMSEA values less than .1, with the fit
535
EXECUTIVE FUNCTIONING
Table 1
Means, Standard Deviations, Standardized Factor Loadings, and Factor Correlations for the 16 Reference Variables in Data Set 1
(N ⫽ 328)
Factor
Variable
Reliability
Age
Gender (% female)
Educationa
Healthb
MMSE
M
50.8
64.0
15.6
2.1
28.8
SD
17.6
Age
correlation
2.6
0.9
1.5
⫺.05
.13
.21*
⫺.18*
Ravens
Shipley Abstraction
Letter Setsc
.84
.74
12.9
11.0
0.78
3.6
2.8
7.4
⫺.37*
⫺.32*
Spatial Relationsd
Paper Foldingc
Form Boardsc
.88
.77
.85
8.8
6.4
7.2
5.2
2.8
4.4
⫺.37*
⫺.30*
⫺.45*
Free Recalle
Paired Associatese
Logical Memoryf
.88
.76
.80
0.7
2.9
45.6
0.1
1.7
9.9
⫺.36*
⫺.36*
⫺.23*
Letter Comparisong
Pattern Comparisong
Digit Symbolh
.87
.90
.93
10.5
16.6
70.9
2.5
3.7
17.5
⫺.46*
⫺.59*
⫺.57*
Synonym Vocabularyi
Antonym Vocabularyi
WAIS Vocabulary
WJ Picture Vocabulary
.83
.82
.90
.86
6.7
6.1
51.5
18.6
3.0
3.0
10.3
5.0
.29*
.17*
.12
.30*
Reasoning
3.5
.84*
.74*
Spatial
visualization
⫺.47*
Episodic
memory
Perceptual
speed
Vocabulary
.83⫹
.90⫹
.84*
.73*
.75*
.76*
.65⫹
.81*
.84*
.82⫹
.87*
.81*
.89*
.80⫹
Factor correlations
Reasoning
Spatial visualization
Episodic memory
Perceptual speed
Vocabulary
—
⫺.48*
.84*
—
.70*
.51*
—
⫺.42*
⫺.44*
.69*
.57*
.57*
—
⫺.66*
.57*
.43*
.38*
.22*
.26*
Note. The plus symbol (⫹) indicates that the unstandardized coefficient was fixed to one to identify the factor. The reliability of the digit symbol variable
was based on the correlation with a parallel form in another sample of 143 adults (Salthouse, Nesselroade, & Berish, 2004). All other reliability estimates
were based on coefficient alpha. The fit statistics were as follows: ␹2(279, N ⫽ 328) ⫽ 94; comparative fit index ⫽ .94; root-mean-square error of
approximation ⫽ .08. MMSE ⫽ Mini-Mental Status Examination; Ravens ⫽ Advanced Progressive Matrices: Set II (Raven, 1962); Shipley ⫽ Shipley
Institute of Living Scale—Revised (Zachary, 1986); WAIS ⫽ Wechsler Adult Intelligence Scale—Third Edition; WJ ⫽ Woodcock–Johnson PsychoEducational Battery—Revised.
a
Number of years of education completed. b Self-rating of health on a scale ranging from 1 (excellent) to 5 ( poor). c Ekstrom et al. (1976). d Bennett,
Seashore, and Wesman (1997). e Salthouse, Fristoe, and Rhee (1996). f Wechsler (1997b). g Salthouse and Babcock (1991). h Wechsler (1997a).
i
Salthouse (1993b).
* p ⬍ .01.
improving as the CFI approaches 1.0 and the RMSEA approaches 0.0. The fit statistics indicate that the five-factor model
provided a relatively good fit to the data, particularly considering
the simple nature of a model in which none of the variables was
related to more than one factor. It should also be noted that all of
the factors had at least moderate correlations with one another, and
each was significantly related to age.
The analyses conducted to investigate the model in Figure 1
added age and the target variable to the correlated cognitive
abilities model. Although the factor loadings and covariances
could have been constrained to the original values when introducing the new variables, they were allowed to vary to evaluate the
robustness of the structure. The rationale is that the structure may
not be very stable if the addition of another variable changes the
pattern of relations among variables and factors. However, in all
cases the addition of age and the target variable to the model
resulted in only minor changes in the structural coefficients relating variables to abilities, and abilities to one another, and thus the
structures can be considered fairly robust.
Parameter estimates for the model in Figure 1 are presented in
Table 2. The second column in the table, labeled Only age,
contains the correlation between age and the target variable that
represents both the direct and indirect influences of age, and the
third column, labeled Unique age, contains the standardized coefficient corresponding to only the direct relation of age on the
variable. Entries in the remaining columns are standardized regres-
536
SALTHOUSE
Table 2
Standardized Coefficients of Relations Between Variables and Age When Considered Alone and Age When Considered in the Context
of Other Cognitive Abilities for Data Set 1
Cognitive ability
Variable
Mini-Mental Status Examination
CLOX1
CLOX2
WCST number categories
WCST proportion correct responses
Letter Fluencya
Category Fluencya
Alternating Fluencya
Connections—Same sequence
Connections—Alternating sequence
Connections—Difference
Only age
Unique age
Reas
Space
Mem
Speed
Voc
⫺.18*
⫺.18*
⫺.14*
⫺.14*
⫺.03
⫺.03
⫺.33*
⫺.33*
⫺.38*
⫺.38*
.05
.05
⫺.19*
⫺.19*
⫺.18*
⫺.18*
⫺.51*
⫺.51*
⫺.17*
⫺.17*
⫺.09
⫺.09
.11
.10
⫺.01
⫺.01
.16
.15
⫺.03
⫺.05
⫺.11
⫺.13
.29*
.31*
.09
.10
.26
.27
⫺.10
⫺.11
.21
.21
.20
.21
.95*
.67*
.84*
.49*
.35
.32
.53*
.47*
.40
.45*
.38
.09
.17
.09
.34
.00
.99*
.58*
.78*
.54*
.65
.47*
⫺.28
—
⫺.37
—
⫺.03
—
⫺.04
—
.07
—
⫺.32
—
⫺.13
—
⫺.35
—
⫺.42*
—
⫺.26
—
⫺.20
—
⫺.05
.00
.05
.11
.02
.03
.05
.06
.09
.08
.06
.11
.18
.20*
.18
.24*
⫺.06
.01
.01
.05
.02
.05
⫺.06
⫺.06
⫺.21
⫺.21
.04
.04
.07
.06
.05
.04
.36*
.36*
.35*
.35*
.53*
.53*
.44*
.44*
.10
.09
.03
.02
⫺.01
.01
.00
.01
.01
.01
.01
.02
.08
.09
.25
.26
.22
.22*
.01
.03
⫺.22
⫺.20
.16
.16
⫺.14
⫺.14
Note. The second row for each variable represents the coefficients when the space construct was not used as a predictor in order to minimize collinearity
with the reasoning construct. Dashes indicate that the predictor variable was not used in the analysis. The cognitive ability factors were based on the
variables described in Table 1 and Figure 1. Reas ⫽ reasoning; Space ⫽ spatial visualization; Mem ⫽ episodic memory; Speed ⫽ perceptual speed; Voc ⫽
vocabulary; CLOX1 ⫽ first part of the CLOX Test (Royall, Corde, & Polk, 1998); CLOX2 ⫽ second part of the CLOX Test; WCST ⫽ Wisconsin Card
Sorting Test.
a
Delis et al. (2001).
* p ⬍ .01.
sion coefficients relating the cognitive ability to the target variable.
Because of the high correlation between the reasoning and spatial
visualization constructs (i.e., .84), some of the estimates could
have been distorted because of collinearity. The analyses were
therefore repeated without the space construct to examine this
possibility. Results from these analyses are presented immediately
below those from the original analyses in Table 2.
The initial target variable was the score on a frequently used
dementia screening instrument, the Mini-Mental Status Examination (Folstein, Folstein, & McHugh, 1975). Entries in the two rows
of Table 2 indicate that this variable was strongly related to
reasoning ability in the current sample of high-functioning adults.
Although there was a small negative correlation between age and
the Mini-Mental Status Examination score, it was eliminated when
age relations through the cognitive abilities were taken into
consideration.
The CLOX Test (Royall, Corde, & Polk, 1998) was designed to
assess aspects of executive functioning in the context of a clock
drawing task. In the first part of the test, CLOX1, the participant is
asked to draw a clock set at a particular time, and in the second
part, CLOX2, he or she simply copies a clock drawn by the
examiner. Both participant drawings are scored in terms of a
number of criteria such as spacing of the numbers, clock hands
represented as arrows with the minute hand longer than the hour
hand, and so forth. The CLOX1 scores had a slight negative
relation with age in this sample, but there was no age relation on
the CLOX2 scores. Only the CLOX1 score was significantly
related to the cognitive abilities, and the influence was restricted to
reasoning ability.
In part because of its widespread usage, the Wisconsin Card
Sorting Test has been characterized as the gold standard for
assessing executive functioning (Delis et al., 2001). The task for
the examinee in this test is to assign cards to categories according
to rules that are changed after every 10 correct responses. There
are many possible measures of performance in this task, but most
are highly correlated with one another in samples of healthy adults
across a wide age range (e.g., Salthouse et al., 2003; Salthouse,
Fristoe, & Rhee, 1996). To illustrate, the correlation between the
number of categories measure and the proportion of correct responses measure in this sample was .87. As one would expect on
the basis of this high correlation, the results of the analyses with
these two measures were very similar. In both cases, the measure
had a strong relation with the reasoning ability, and there were no
unique age-related effects after considering influences of age
through the cognitive abilities.
The Letter Fluency, Category Fluency, and Alternating Fluency
tasks (Delis et al., 2001) all involve the examinee generating as
many words as possible within a limited time that satisfy specified
constraints. In the Letter Fluency task, the constraint is that the
words should begin with a particular letter of the alphabet; in the
Category Fluency task, the constraint is that the words should
belong to a particular category; and in the Alternating Fluency
537
EXECUTIVE FUNCTIONING
task, the constraint is that every other word should belong to a
different one of two designated categories. These types of tasks are
often assumed to involve maintenance of a set and inhibition of
past responses.
Each of the fluency measures had a significant relation with
perceptual speed ability, and there were also a few significant
relations with the vocabulary and memory abilities. It is interesting
that the relations with age were reversed in direction after controlling the influences through the cognitive abilities, and the adjusted
relation was statistically significant for the Letter Fluency measure.
The Connections Test (Salthouse et al., 2000) is a variant of the
Trail Making Test in which the test pages consist of adjacent
circles containing either numbers or letters, and the four conditions
involve numeric, alphabetic, alternating numeric and alphabetic,
and alternating alphabetic and numeric sequences. As in the Trail
Making Test, the task for the participant is to draw a line connecting the items in the appropriate sequence as rapidly as possible.
The times to connect all of the items in the two simple sequences
(i.e., numbers only and letters only) are averaged to form a same
sequence measure that is roughly equivalent to Version A of the
Trail Making Test, and the two times in the alternating sequences
are averaged to form an alternating sequence measure that is
roughly equivalent to Version B of the Trail Making Test.
The results in Table 2 indicate that the pattern was somewhat
different across the three connections measures. However, each
had a strong relation to reasoning ability, and in none of them was
there a significant direct relation to age after considering the
influence through other cognitive abilities. Only the same sequence
connections measure had a significant relation to perceptual speed
ability.
There are two major findings from the results summarized in
Table 2. First, although most of the measures were negatively
related to age, very few of the direct age relations, which were
statistically independent of age-related influences through the cognitive abilities, were significantly different from zero. Second, all
of the measures had significant relations with one or more cognitive abilities, and most were influenced by either reasoning ability
or perceptual speed ability. As noted in the introduction, this is not
the pattern of results one would expect if the measures represented
a distinct executive functioning construct that has unique relations
to age.
Data Set 2
The major purpose of the analyses of the second data set was to
replicate the analyses based on the model in Figure 1 with a more
extensive set of target variables. The larger collection of variables
hypothesized to reflect executive functioning was made possible
by aggregating data across many previous studies. The aggregate
data set has a great deal of missing data, but the missing values are
due to the study to which the individual was assigned rather than
to the level of the variable, and thus the data can be assumed to be
missing at random. This property allows powerful analytical procedures to be used to deal with missing data, such as maximumlikelihood estimation (Arbuckle, 1996; Enders & Bandalos, 2001).
Additional information about the aggregate data set and the analytical procedure is available in another report that focused on
different research questions (Salthouse, 2004).
Method
The analyses were based on data from a total of 6,959 different individuals from 34 studies conducted in Timothy A. Salthouse’s laboratory,
excluding the study described above. All of the studies were similar in that
the participants in a given study were administered the tasks in the same
order, and the samples consisted of adults spanning a continuous distribution of ages between 18 and 95. Studies with only two extreme age groups
(e.g., younger and older adults) and those in which the younger adults were
primarily college students were excluded to avoid inflating the estimates of
the age relations when the variance associated with middle-aged adults is
omitted, and to minimize confounding age with student status.
Only a subset of the 16 reference cognitive variables was included in
most of the studies, but each study contained at least two of the reference
variables, in addition to information about age and other demographic
characteristics of the participants. Furthermore, each reference variable
was assessed in at least three studies with different combinations of other
variables. Many of the studies also included measures that have been
hypothesized to represent executive functioning, which allowed examination of their relations with the cognitive abilities according to the model
portrayed in Figure 1. More details on the target variables and the characteristics of the samples are available in the specific studies, which are
cited in Table 4.
As in Data Set 1, the model in Figure 1 was investigated with the AMOS
(Arbuckle, 2003) structural equation program. This program incorporates a
maximum-likelihood estimation algorithm to take advantage of all of the
available information at the time of analysis and thus is an efficient method
of dealing with incomplete or missing data (e.g., Arbuckle, 1996; Enders &
Bandalos, 2001).
Results and Discussion
Table 3 contains summary statistics for the reference cognitive
variables, including factor loadings and factor correlations, derived
from a confirmatory factor analysis specifying five correlated
factors. Comparison of the results in Tables 1 and 3 reveals that the
variable means, the factor loadings, and the factor correlations
were very similar in the two independent data sets.
Because of the large number of variables in the aggregate data
set, several additional analyses were conducted to explore the
generalizability of the analytical procedure based on the model in
Figure 1. One analysis focused on a target variable consisting of
the number of words answered correctly on a New York Times
crossword puzzle with a time limit of 15 min. Results with this
measure are summarized in the top row of Table 4. The pattern of
standardized coefficients suggests that crossword puzzle proficiency is primarily influenced by vocabulary ability, with small
and not statistically significant influences of reasoning and memory abilities. Furthermore, to account for the age-related effects
observed on this measure of crossword puzzle performance, the
results imply that one must account for effects operating through
vocabulary ability and also for positive age-related effects on at
least one additional factor that is associated with better crossword
puzzle performance. The analysis does not indicate the nature of
the other factor(s) responsible for any direct or unique age-related
influences, but in this particular case it may be related to a greater
accumulation of a specific type of knowledge required in these
crossword puzzles.
A second set of analyses was conducted on new variables
hypothesized to be primarily influenced by one of the five cognitive abilities. Results with these variables serve as a validity check
on the analytical method because to the extent that the procedure
538
SALTHOUSE
Table 3
Means, Standard Deviations, Age Correlations, and Standardized Factor Loadings for the Reference Variables
Factor
N
M
SD
Age
correlation
Age
Gender (% female)
Educationa
Healthb
6,959
6,959
5,208
6,943
49.1
58.0
15.3
2.1
17.2
4.9
2.6
1.0
—
.02
⫺.01
.13*
Ravens
Shipley Abstraction
Letter Setsc
2,110
1,408
1,308
7.6
13.2
10.0
3.7
4.6
3.9
⫺.50*
⫺.30*
⫺.26*
Spatial Relationsd
Paper Foldingc
Form Boardsc
1,290
1,129
990
9.3
6.5
7.2
5.1
2.9
4.2
⫺.36*
⫺.44*
⫺.41*
Free Recalle
Paired Associatese
Logical Memoryf
1,895
1,897
923
0.7
2.8
45.1
0.2
1.7
10.2
⫺.41*
⫺.37*
⫺.25*
Letter Comparisong
Pattern Comparisong
Digit Symbolh
6,212
6,212
2,177
10.4
16.4
77.1
3.2
4.2
19.1
⫺.43*
⫺.52*
⫺.57*
Synonym Vocabularyi
Antonym Vocabularyi
WAIS Vocabulary
WJ Picture Vocabulary
3,640
3,637
927
930
6.9
6.3
51.4
19.1
2.9
3.2
10.3
5.4
.28*
.18*
.09*
.29*
Variable
Reasoning
Spatial
visualization
Episodic
memory
Perceptual
speed
Vocabulary
.89⫹
.86*
.80*
.91⫹
.83*
.79*
.79*
.73*
.74⫹
.80*
.82*
.78⫹
.89*
.90*
.85*
.77⫹
Factor correlations
Reasoning
Spatial visualization
Episodic memory
Perceptual speed
Vocabulary
—
⫺.47*
.92*
—
.73*
.66*
—
⫺.40*
⫺.47*
.78*
.66*
.70*
—
⫺.61*
.45*
.43*
.47*
.29*
.27*
Note. The plus symbol (⫹) indicates that the unstandardized coefficient was fixed to one to identify the factor. The fit statistics were as follows: ␹2(753) ⫽
94; comparative fit index ⫽ .96; root-mean-square error of approximation ⫽ .03. Ravens ⫽ Advanced Progressive Matrices: Set II (Raven, 1962); Shipley
⫽ Shipley Institute of Living Scale—Revised (Zachary, 1986); WAIS ⫽ Wechsler Adult Intelligence Scale—Third Edition; WJ ⫽ Woodcock–Johnson
Psycho-Educational Battery—Revised.
a
Number of years of education completed. b Self-rating of health on a scale ranging from 1 (excellent) to 5 ( poor). c Ekstrom et al. (1976). d Bennett,
Seashore, and Wesman (1997). e Salthouse, Fristoe, and Rhee (1996). f Wechsler (1997b). g Salthouse and Babcock (1991). h Wechsler (1997a).
i
Salthouse (1993b).
* p ⬍ .01.
is valid, these variables would be expected to have strong relations
with the relevant ability but relatively weak relations with other
abilities.
Examination of the relevant rows in Table 4 indicates that
measures postulated to represent different cognitive abilities generally had their highest loadings on the hypothesized ability. The
Surface Development (Ekstrom, French, Harman, & Dermen,
1976) and Word Recall (Wechsler, 1997b) measures still had some
unique age-related influences after considering influences of age
through other abilities, which implies that not all of the age-related
effects on these particular measures were shared with age-related
effects on the cognitive abilities represented in the analysis. However, for the most part, the pattern in the top of Table 4 is
consistent with the interpretation that the measures are primarily
influenced by the hypothesized abilities and that most of the
age-related effects on the measures are shared with age-related
effects on the hypothesized ability and on other cognitive abilities.
As noted earlier, the strong correlation between the reasoning
and space ability constructs (i.e., .92 in Table 3) can distort
estimates of the influences of these abilities because of collinearity
when both predictors are considered simultaneously. The analyses
of the reasoning and spatial visualization measures were therefore
repeated after dropping either the space or the reasoning construct
from the model. For the Series Completion (Salthouse & Prill,
1987) and Analysis Synthesis (Woodcock & Johnson, 1990)
measures, deletion of the space construct resulted in significant
loadings only on the reasoning construct (i.e., .56 for Series
Completion and .76 for Analysis Synthesis). For the Block Design
(Wechsler, 1997a) measure, deletion of the reasoning construct
resulted in a loading of .78 on the space construct and no other
significant loadings, and for the Surface Development measure,
there were only significant loadings of .75 on the spatial visualization ability and .32 on the vocabulary ability after the reasoning
construct was deleted.
EXECUTIVE FUNCTIONING
The results just described suggest that the analytical procedure
based on the model in Figure 1 is informative about the relative
magnitude of relations between different cognitive abilities and the
target variable and about the unique relation of age on the variable
after controlling age-related influences on several cognitive abilities. The procedure was therefore applied to measures often hypothesized to reflect executive functioning or aspects of executive
control.
The first set of variables under the Executive functioning heading in Table 4 are the same variables examined in Data Set 1.
Comparison of the results in Tables 2 and 4 indicates that the
pattern was very similar in the two independent data sets. Specifically, the results in Table 4 indicate that the number of categories
measure of the Wisconsin Card Sorting Test performance was
strongly related to reasoning ability (and significantly so when the
space construct was deleted from the analysis) and that there were
no unique age-related effects on the number of categories measure
after considering age-related influences through the set of cognitive abilities. Moreover, the fluency measures were all influenced
by speed ability, and in each of these measures, the age relation
was reversed in direction from negative to positive when the
influences through the cognitive abilities were taken into consideration. Finally, the same-sequence connections measure was
again influenced by reasoning and speed abilities, with the other
connections measures having a mixed pattern of influences, and
there were no unique age-related effects on any of the measures
after considering influences through the cognitive abilities.
The Ruff Figural Fluency Test (Ruff, 1996) involves the participant attempting to draw lines between dots to generate as many
different figures as possible within a fixed period of time. It is
therefore similar to the Verbal Fluency tests but without any
involvement of words. It can be seen in Table 4 that this measure
was significantly related to perceptual speed ability and that there
were no unique age-related effects on the measure after considering influences through the cognitive abilities.
The Tower of Hanoi task requires the participant to change one
arrangement of disks on three towers to a different arrangement in
the fewest number of moves, with the constraints that only one
disk can be moved at a time and that a larger disk can never be
placed on top of a smaller disk. Performance on this task is
sometimes postulated to assess planning and goal management
aspects of executive functioning. The results in Table 4 indicate
that the average number of moves required in the Tower of Hanoi
task was related to reasoning and memory abilities, but only the
memory relation was statistically significant because of the large
standard error for the reasoning relation. Both the influences from
reasoning and from memory were significant when the space
construct was deleted from the analysis. It is important to note that
there were no unique age-related effects on the number of moves
in the Tower of Hanoi task when age-related influences through
the cognitive abilities were taken into consideration. This pattern
implies that at least with respect to individual differences associated with aging, the Tower of Hanoi task does not seem to measure
anything distinct from what is measured with the variables used to
represent the five cognitive abilities in the model portrayed in
Figure 1.
The Sort Recognition Test is included in the Delis–Kaplan
Executive Function System (Delis et al., 2001) to assess conceptual reasoning ability. The task for the participant is to identify the
basis for grouping objects into two categories. The number of
539
correct classifications in this test was weakly and nonsignificantly
related to the reasoning, space, and memory abilities. After deletion of the space construct, the influence of reasoning ability was
larger and statistically significant. There were no unique agerelated effects on the Sort Recognition measure after considering
indirect age-related effects through the cognitive abilities.
The Proverb Interpretation test is a subtest included in the
Delis–Kaplan Executive Function System (Delis et al., 2001) to
assess verbal abstraction skills. The version of the test used in this
project required the participant to select the best interpretation of
the proverb from a set of alternatives. Performance on this test was
positively related to reasoning ability and vocabulary ability but
was negatively related to space ability. When the space construct
was deleted from the analysis, none of the relations with any of the
cognitive abilities were significant.
The Trail Making Test requires the examinee to draw lines to
connect circles in a specified sequence as rapidly as possible. In
Version A of the test, the circles contain numbers that are to be
connected in numerical sequence. In Version B, the circles contain
both numbers and letters, and the task is to connect the items in
alternating numeric and alphabetic sequence. Performance in B
relative to A is often assumed to reflect planning, switching, and
maintaining the current status in each of the two sequences. No
spatial visualization variables were included in the samples in
which the Trail Making Test was administered, and thus it was not
possible to examine the influence of the space ability construct on
these measures. The Trail Making A, B, and B – A measures were
all primarily influenced by speed, although the B-minus-A difference was also influenced by reasoning. There were no unique age
effects in any of the Trail Making measures when age-related
influences through the cognitive abilities were controlled.
Switching tasks have become popular as a means of investigating the efficiency of controlling one’s attention. The versions used
in the study included in the data set were reaction time tasks in
which initially the participant was to respond with one rule (e.g.,
press the right key for an odd digit and press the left key for an
even digit), but when a signal occurred he or she was to switch to
a different rule (e.g., press the right key for a digit greater than five
and press the left key for a digit less than five). The primary
measure of performance in these tasks is the difference between
reaction time in the preswitch trials and reaction time on the trial
immediately after the switch signal. The values in Table 4 indicate
that the pattern of relations with cognitive abilities was inconsistent across the three measures with different combinations of rules,
but in none of the measures was there a significant unique relation
from age.
One aspect of cognitive control that has frequently been mentioned as a component of executive functioning is inhibition of
prepotent responses. A popular test of this type of inhibition is the
Stroop Color–Word Test in which the task is to name the colors of
letter strings when the strings are XX’s (neutral), when they are
names of the colors in which they are written (congruent), and
when they are names of colors different from those in which they
are written (incongruent). Naming times in this type of situation
are typically very slow in the incongruent condition, and it is often
assumed that this occurs because two potential responses (i.e.,
word name and color name) are in conflict and some type of
inhibition is needed to suppress the currently irrelevant response.
All of the Stroop measures, including the incongruent-minuscongruent difference score, were primarily influenced by speed
540
SALTHOUSE
Table 4
Standardized Coefficients of Relations Between Variables and Age When Considered Alone and Age When Considered in the Context
of Other Cognitive Abilities
Cognitive ability
Unique
age
Reas
Space
Mem
Speed
.47*
.21
.02
.15
⫺.06
⫺.17
⫺.07
.11
.12
.25
.01
.09
.07
.21
.06
.18
⫺.02
.12
.08
⫺.01
⫺.08
⫺.11
.02
Variable
Studies
New York Times crossword puzzle
Reasoning
Series Completiona
WJ Analysis Synthesis
Space
Block Designb
Surface Developmentc
Memory
New Recall (List B)d
Word Recalld
Speed
Cross Oute
Digit Digit RTf
Vocabulary-Knowledge
Multiple-Choice Knowledgeg
Short Answer Knowledgeg
Executive functioning
WCST Number Categories
Letter Fluencyh
Category Fluencyh
Alternating Fluencyh
Connections—Same sequence
Connections—Alt. sequence
Connections—Difference
Ruff Figural Fluencyi
Tower of Hanoij
Sort Recognitionh
Proverb Interpretationh
Trail Making A
Trail Making B
Trail Making B ⫺ A difference
Switch 1
Switch 2
Switch 3
Inhibition
Stroop congruent
Stroop neutral (N)
Stroop incongruent (I)
Stroop I ⫺ N difference
Reading time normal
Reading time distraction
Reading time difference
Reading acc normal
Reading acc distraction
Reading acc difference
Anti-Cue time congruent
Anti-Cue time incongruent
Anti-Cue time difference
Anti-Cue acc congruent
Anti-Cue acc incongruent
Anti-Cue acc difference
Working Memory
WAIS Letter-Number Sequence
Computation Spank
Reading Spank
NBack-0j
NBack-1j
NBack-2i
Keeping Trackj
Matrix Monitoring 1j
25, 26, 30
601
.43*
31
31
150
204
⫺.37*
⫺.36*
.12
.09
.64
.72
16, 31
1, 2, 30
463
639
⫺.39*
⫺.32*
.00
⫺.21*
⫺.73*
⫺.88*
16, 31–34
33
1,203
329
⫺.36*
⫺.50*
.09
⫺.22*
⫺.15
.18
.07
⫺.21
.94*
.74*
31
8–14, 17–22, 27
204
2,417
⫺.71*
⫺.41*
⫺.10
⫺.03
.03
⫺.08
.09
.01
⫺.13
.34*
⫺.20
⫺.07
.37
.22
.23
.23
.02
⫺.07
23–26, 30
25, 30
16, 18, 32
16, 19, 23, 24, 31–33
19, 33
33
27–34
27–34
27–34
32
32
33
33
16, 19
16, 19
16, 19
20
20
20
14,
14,
14,
14,
15,
15,
15,
15,
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
32
31
3–6, 12, 14, 18
3–6, 12, 14, 18
32
32
32
32
32
n
Only
age
803
401
⫺.05
.09
.10
.19*
1.21*
1.48*
.85*
.41*
Voc
.47*
⫺.04
⫺.08
.60*
.60*
711
1,447
454
330
1,520
1,520
1,520
261
254
330
330
383
383
383
159
159
159
⫺.28*
⫺.13*
⫺.19*
⫺.20*
⫺.41*
⫺.21*
⫺.14*
⫺.44*
⫺.29*
⫺.45*
⫺.21*
⫺.50*
⫺.55*
⫺.47*
⫺.37*
⫺.37*
⫺.42*
⫺.12
.15*
.33*
.22
.03
.08
.07
⫺.01
.08
⫺.11
⫺.06
⫺.28
⫺.10
.00
⫺.06
.02
⫺.22
.62
.19
⫺.08
⫺.42
.47*
.07
.02
.31
.52
.28
.72*
.22
.18
.34
.13
.12
⫺.23
⫺.09
⫺.06
.03
.19
⫺.15
⫺.05
.04
.19
.05
.26
⫺.65*
—
—
—
—
—
—
⫺.17
.14
.50*
.43*
.03
.05
.05
⫺.04
.34*
.21
.11
⫺.27
⫺.01
.11
.31
.32
.21
.01
.32*
.51*
.54*
.40*
.21*
.13
.40*
⫺.17
⫺.02
.12
.78*
.55*
.34*
.25
.21
.42
.12
.21*
⫺.03
⫺.11
⫺.08
⫺.12
⫺.13
⫺.01
⫺.27*
.05
.20
.25
⫺.01
⫺.13
⫺.22
⫺.01
.10
681
681
681
681
260
261
261
261
261
261
255
255
255
255
255
255
⫺.41*
⫺.44*
⫺.57*
⫺.48*
⫺.19*
⫺.18*
⫺.14
⫺.23*
⫺.20*
.05
⫺.45*
⫺.49*
⫺.25*
⫺.27*
⫺.39*
.18*
.00
.06
.01
⫺.16
⫺.22
⫺.07
⫺.05
⫺.17
⫺.02
⫺.07
⫺.24
⫺.37*
⫺.32*
⫺.06
⫺.15
.13
.32
.13
.53
.26
⫺.14
.33
.32
.35
.51
⫺.29
.01
⫺.01
⫺.04
.11
.20
⫺.13
⫺.31
⫺.23
⫺.25
⫺.03
⫺.25
⫺.28
⫺.24
⫺.19
⫺.04
⫺.06
⫺.16
⫺.12
.02
.06
.10
⫺.07
⫺.01
.02
.11
.07
.05
⫺.02
⫺.03
.27*
.38*
⫺.22
.04
⫺.12
⫺.28*
⫺.01
⫺.18
.23
.57*
.74*
.48*
.25*
.48*
.38*
.34*
⫺.30
⫺.31*
.13
.41*
.42*
.17
.24
.35*
⫺.18
⫺.15
⫺.04
⫺.12
⫺.02
.51*
.38*
.36*
⫺.08
.02
⫺.06
.05
.07
.20
⫺.05
.02
⫺.08
204
1,461
1,461
261
261
261
257
258
⫺.40*
⫺.29*
⫺.36*
⫺.24*
⫺.36*
⫺.38*
⫺.23*
⫺.17*
⫺.03
.07
⫺.23
⫺.04
⫺.02
⫺.02
⫺.10
.12
⫺.16
—
—
⫺.13
⫺.38
⫺.07
⫺.31
⫺.05
.23
—
—
.26
.20
.09
.07
.06
.15
⫺.22*
.34*
⫺.00
.25
.09
⫺.19
.08
.07
⫺.18
.33*
⫺.15
⫺.14
⫺.09
.04
⫺.14
.40
.95*
.00
.22
.69*
.72*
.83*
.48
541
EXECUTIVE FUNCTIONING
Table 4 (continued)
Cognitive ability
Variable
Working Memory (continued)
Matrix Monitoring 2j
Running Memoryl 1-back
Running Memory 2-back
Running Memory 3-back
Studies
n
Only
age
32
33
33
33
257
330
330
330
⫺.26*
⫺.34*
⫺.27*
⫺.20*
Unique
age
Reas
Space
Mem
Speed
Voc
.20
.01
.19
⫺.05
.95*
⫺.11
.60*
.10
⫺.17
.17
⫺.22
.12
.07
.24
.14
.22
.08
.19
.30*
.17
⫺.18
.12
.00
.08
Note. All target variables expressed in time or errors were recoded such that a higher score represented better performance, and a negative correlation
with age reflected poorer performance with increased age. Dashes indicate that data were not obtained. Reas ⫽ reasoning; Space ⫽ spatial visualization;
Mem ⫽ episodic memory; Speed ⫽ perceptual speed; Voc ⫽ vocabulary; 23, 24, 25, 26 ⫽ Hambrick, Salthouse, and Meinz (1999); 30 ⫽ Salthouse
(2001b); 31 ⫽ Salthouse and Ferrer-Caja (2003); 16 ⫽ Salthouse, Fristoe, and Rhee (1996); 1 ⫽ Salthouse and Mitchell (1990); 2 ⫽ Salthouse, Babcock,
Skovronek, Mitchell, and Palmon (1990); 32 ⫽ Salthouse, Atkinson, and Berish (2003); 33 ⫽ Salthouse, Berish, and Siedlecki (2004); 34 ⫽ Salthouse,
Nesselroade, and Berish (2004); 8, 9 ⫽ Salthouse (1994b); 10, 11 ⫽ Salthouse (1994a); 12 ⫽ Salthouse (1995); 13 ⫽ Salthouse, Fristoe, Lineweaver, and
Coon (1995); 14 ⫽ Salthouse and Meinz (1995); 17 ⫽ Salthouse, Hambrick, Lukas, and Dell (1996); 18 ⫽ Salthouse, Hancock, Meinz, and Hambrick
(1996); 19 ⫽ Salthouse, Toth, Hancock, and Woodard (1997); 20 ⫽ Salthouse, Fristoe, McGuthry, and Hambrick (1998); 21 ⫽ Meinz and Salthouse
(1998); 22 ⫽ Salthouse, McGuthry, and Hambrick (1999); 27 ⫽ Salthouse et al. (2000); 28, 29 ⫽ Salthouse (2001a); 15 ⫽ Salthouse (1996); 3 ⫽ Salthouse
and Babcock (1991); 4, 5 ⫽ Salthouse (1992); 6 ⫽ Salthouse (1993a); WJ ⫽ Woodcock–Johnson Psycho-Educational Battery—Revised; RT ⫽ reaction
time; WCST ⫽ Wisconsin Card Sorting Test; Alt. ⫽ alternating; Trail Making A ⫽ Trail Making Test, Version A (Barnes et al., 2003); Trail Making B ⫽
Trail Making Test, Version B; Stroop ⫽ Stroop Color Word Test (Barnes et al., 2003); acc ⫽ accuracy; WAIS ⫽ Wechsler Adult Intelligence Scale—Third
Edition.
a
Salthouse and Prill (1987). b Wechsler (1997a). c Ekstrom et al. (1976). d Wechsler (1997b). e Woodcock and Johnson (1990). f Salthouse
(1992). g Hambrick, Salthouse, and Meinz (1999). h Delis et al. (2001). i Ruff (1996). j Salthouse et al. (2003). k Salthouse and Babcock
(1991). l Siedlecki, Salthouse, and Berish (2005).
* p ⬍ .01.
ability, with the incongruent measure also significantly related to
reasoning ability. After controlling for the age-related influences
on the cognitive abilities, it was found that none of the Stroop
measures had unique age-related effects.
Several of the tasks used to evaluate inhibition assessed performance both in terms of time and accuracy, and thus both measures
are considered separately in these analyses. (Relations of cognitive
abilities on composite measures of time and accuracy were examined with data from a single study in Salthouse et al., 2003.)
The Reading With Distraction task (Salthouse et al., 2003)
resembles the Stroop task in that in order to be successful the
participant must inhibit the tendency to make an irrelevant response. The task involves the participant reading short passages
and answering questions about them in two conditions. In one
condition, the to-be-read text is normal, and in the other condition,
distracting words in a different font are interleaved within the
normal text. The primary performance measures in this task are the
reading time in the normal condition, the reading time in the
distraction condition, and the difference between normal reading
time and reading time with distraction. It can be seen in Table 4
that all of the reading time measures were primarily influenced by
the speed and vocabulary abilities. There were no significant
unique age effects on the reading time measures after adjusting for
influences through the cognitive abilities.
Comprehension accuracy in the Reading With Distraction task
was assessed in terms of accuracy of answering questions about
the passage immediately after reading it. The reading accuracy
measures were mostly influenced by memory ability, but they also
had a negative relation to speed ability. There were no unique
age-related effects on the reading accuracy measures after controlling age-related effects on the cognitive abilities.
The Anti-Cue task (Salthouse et al., 2003) was based on the
Antisaccade task (de Jong, 2001) that has been used to investigate
reflexive shifts of attention. In the version used in the study
included in the aggregate data set, the respondent fixates in the
middle of the computer screen, a flash (cue) is briefly presented on
the left or right side of the display, and then a target letter (X or O)
is presented either on the same side as the cue (congruent) or on
the opposite side (incongruent). Both time and accuracy of the
decisions are monitored in each condition. The decision time
measures in this task were influenced by speed ability, but there
were also strong unique age-related effects which indicate that the
age-associated influences on these measures were not fully captured by the age-related influences on the cognitive abilities. The
accuracy measures in the Anti-Cue task had weak and inconsistent
relations to the cognitive abilities, but there were no unique agerelated effects after controlling the influences of age on those
abilities.
Several target variables were designed to assess simultaneous
processing and storage aspects of working memory, but the results
from the analyses were quite complex. In the Letter-Number
Sequencing task from the Wechsler Adult Intelligence Scale—
Third Edition (Wechsler, 1997a), the participants listened to an
intermixed sequence of letters and digits and then reported the
numbers in numerical order followed by the letters in alphabetical
order. This measure did not have significant relations to any of the
cognitive abilities, but it also had no unique age-related effects.
The Computation Span and Listening Span tasks (Salthouse &
Babcock, 1991) were similar to the letter sequencing task in that
the examinee attempted to carry out specified processing while
keeping other information in memory. In the Computation Span
task, the participant viewed a series of simple arithmetic problems
that were to be solved by selecting the correct answer from a set of
alternatives and also tried to remember the last digit in each
problem for later recall. In the Listening Span task, the participant
heard a series of sentences and then answered a question about the
sentence while trying to remember the last word in each sentence
for later recall. Although these two tasks were designed to be very
542
SALTHOUSE
similar and were administered to the same individuals, the measures had different patterns of relations with the cognitive abilities.
For example, the Computation Span measure was positively related to reasoning and negatively related to speed, whereas the
Listening Span measure was positively related to speed and to
vocabulary. However, there was no unique age-related influence
on either measure after considering age-related influences through
the cognitive abilities. The inconsistency in the patterns may be at
least partially attributable to the weak representation of different
cognitive abilities in the studies in which these tasks were administered and in particular to the absence of any samples in which
reference variables from space or memory abilities were assessed.
The N-back task (Salthouse et al., 2003) involves the research
participant listening to or viewing a sequence of digits and reporting the items that occurred N items back in the sequence. Versions
of this task have been popular in neuroimaging studies because the
stimuli and responses can remain the same while the processing
load is increased by requiring items farther back in the sequence to
be reported. Inspection of Table 4 indicates that performance in
this task was related to reasoning ability, with a stronger influence
when the presumed mental demands increased from the 0-back to
either the 1-back or 2-back condition. In none of the measures
were there significant unique age-related effects after controlling
age-related effects on the cognitive abilities.
The Keeping Track task (Salthouse et al., 2003) involves the
presentation of a sequence of words from different categories, with
the sequence periodically interrupted by queries in which the
research participant was to determine whether the displayed word
was the most recent member of the designated category. Performance on this task was strongly related to reasoning ability, and
there were no significant unique age-related effects after considering effects through other cognitive abilities.
The Matrix Monitoring task (Salthouse et al., 2003) consisted of
a series of four displays. The first contained a matrix with a dot in
one cell, and the next two contained arrows indicating the direction
in which the dot should be moved. The final display portrayed a
dot in a new position in the matrix, with the research participant
asked to decide whether this was the correct position for the dot on
the basis of the initial position and the sequence of movements
indicated by the arrows. There were two conditions in this task, the
one-matrix condition just described and a two-matrix condition in
which the same type of monitoring had to be carried out simultaneously with two matrices. Matrix monitoring performance was
related to reasoning ability, with a stronger relation in the more
demanding two-matrix condition. Deletion of the space construct
from the analysis with the two-matrix measure resulted in the
relation with reasoning remaining strong (i.e., .94), but deletion of
the reasoning construct resulted in a much smaller relation with the
space construct (i.e., .37). In none of the analyses with the Matrix
Monitoring measures were the direct age-related effects significantly different from zero after controlling the influences of age on
the cognitive abilities.
The Running Memory task (Siedlecki, Salthouse, & Berish,
2005) involved the examinee viewing a sequence of words and
then occasionally being asked to recall the last three words in the
sequence. Performance was assessed as the percentage of words
correctly recalled in each serial position. Examination of Table 4
indicates that there was an inconsistent mixture of influences
across the three serial positions, as the influence of reasoning
ability was greater in the second position than in the first (most
recent) position but was smaller in the third position. After controlling for age-related influences on the cognitive abilities, it was
discovered that there were no unique age relations on the Running
Memory measures.
General Discussion
The research reported in this article used the analytical model
portrayed in Figure 1 to investigate the meaning of variables
frequently hypothesized to reflect executive functioning. The rationale was that if the target variables represent something different from the cognitive abilities included in the model, then the
variables not only should have relatively weak relations to those
abilities but also should have significant unique (direct) relations
with an individual-difference variable such as age if they are
reliably influenced by another construct, such as executive functioning, that is related to age.
Evidence for the plausibility of the analytical method is available from the findings that measures hypothesized to reflect different cognitive abilities had expected patterns of loadings on their
respective abilities and had few unique age-related influences.
Moreover, the method is capable of identifying variables in which
there are strong direct relations from age because the crossword
puzzle performance measure had a substantial unique relation with
age after considering age-related influences through the cognitive
abilities. This pattern is consistent with the possibility that crossword puzzle performance is influenced by a construct not represented by the cognitive abilities in the model, which in this case
probably corresponds to a specialized type of knowledge that
increases with age among these individuals.
The similar pattern of results with two quite different data sets
also indicates that the analytical method yields replicable results.
One data set was based on a modest-sized sample of 328 adults
with complete data on all reference variables but only a few
variables postulated to reflect executive functioning, and the other
was based on data from nearly 7,000 adults aggregated across 34
different studies with a large proportion of missing values for the
reference variables but a substantial number of candidate executive
variables.
Application of the analytical method to variables hypothesized
to reflect the neuropsychological construct of executive functioning revealed that many of the target variables are closely related to
the cognitive abilities of reasoning and perceptual speed. Furthermore, this was equally true for measures derived from traditional
neuropsychological tests as for measures intended to assess specific aspects of cognitive control such as inhibition or working
memory. Another pattern apparent in Tables 2 and 4 was that very
few of the executive functioning variables had age-related influences that were statistically independent of the age-related influences on the cognitive abilities. Because this indicates that most of
the variance shared between the target variable and age overlapped
with the variance in established cognitive abilities, it is inconsistent with the interpretation that the target variable represents a
distinct construct. The outcomes of these analyses therefore suggest that measures currently used by neuropsychologists to assess
executive functioning may not represent novel aspects of functioning in normal adults. Instead they appear to reflect the same
dimensions of differences assessed by more traditional cognitive
tests that may have superior psychometric properties (in terms of
reliability and sensitivity).
543
EXECUTIVE FUNCTIONING
Although the results summarized in Tables 2 and 4 appear
convincing, as with any project there are several limitations. One
is that only five cognitive abilities were represented in the analyses, and the number of conceptually distinct cognitive abilities
could be as many as 40 or more (Carroll, 1993). Furthermore, the
representation of even these five abilities was not complete for all
of the target variables because reference variables for some abilities were not included in several studies in the second data set,
which precluded analyses of the relations between the ability and
the target variable.
A second limitation of the analyses is that it is not clear whether
the different pattern of results across ostensibly similar measures
reflects weaknesses of the analytic method or unexpected complexity of the measures. The inconsistent pattern is particularly
apparent with the switching measures used to assess efficiency of
redirecting attention and with the working memory span measures
used to assess working memory. In both cases, very similar target
measures were obtained from the same participants and with the
same combinations of reference cognitive variables, and yet the
measures differed in their patterns of relations with cognitive
abilities. These inconsistencies may reflect the involvement of
construct-irrelevant characteristics such as the type of stimulus
material (i.e., words vs. numbers) or the particular processing
requirements (e.g., listening or reading vs. performing simple
arithmetic). Until the reasons for the different patterns of relations
with cognitive abilities are understood, however, researchers
should be cautious in concluding that measures represent the same
construct and in inferring that they would have the same patterns
of relations with other constructs merely because they appear to
involve similar processes.
Finally, although Figure 1 represents cognitive abilities as influencing the target variables, it is important to recognize that the
direction of the relation between abilities and the target variable
depends on one’s theoretical perspective and cannot be distinguished on the basis of the current analyses. That is, latent constructs in the analyses conducted here correspond to variance that
is common across variables, but the analyses are not necessarily
informative about the reasons for that commonality. The latent
constructs could correspond to abilities that affect the efficiency or
effectiveness of the processes required in the tasks (as is assumed
in the representation in Figure 1), or they could reflect the operation of critical processes that are involved in each of the measures
postulated to assess different cognitive abilities (in which case the
arrows in Figure 1 might be from the target variable to the ability
constructs). Miyake and colleagues (e.g., Friedman & Miyake,
2004; Miyake et al., 2000) have adopted this latter type of process
interpretation as they have tried to identify latent constructs that
represent different theoretical processes and have then examined
the relations of these constructs to performance on various cognitive or neuropsychological tests. Although the two approaches
differ in how the constructs are conceptualized and in which
theoretical concepts are assumed to be most fundamental, both
acknowledge that there are strong relations among neuropsychological and cognitive variables and even stronger relations among
latent constructs representing the variance shared among variables.
Another common theme in the two approaches is the assumption
that there are seldom one-to-one relations between variables and
theoretical constructs and hence that few variables can be considered to be pure, either with respect to processes or to abilities.
In conclusion, this article raised the question of the type of
evidence needed to establish that measures postulated to assess
executive functioning actually do represent that construct. It was
proposed that a potentially informative method of investigating the
meaning of a variable is to examine the pattern of relations it has
with established cognitive abilities and the degree to which the
variable has unique relations with an individual-difference variable
such as age. The results of the analyses suggest that many variables
postulated to assess executive functioning in healthy adults are
closely related to cognitive abilities of reasoning and perceptual
speed. Furthermore, although nearly all of the purported executive
functioning variables were negatively related to age, only a small
proportion of the age-related influences were statistically independent of the age-related influences on the cognitive abilities. Taken
together with the recent findings of Salthouse et al. (2003), these
results raise questions about the extent to which executive functioning represents a distinct dimension of individual differences in
healthy adults.
References
Arbuckle, J. L. (1996). Full information estimation in the presence of
incomplete data. In G. A. Marcoulides & R. E. Schumacker (Eds.),
Advanced structural equation modeling (pp. 243–277). Mahwah, NJ:
Erlbaum.
Arbuckle, J. L. (2003). AMOS for Windows [Computer software]. Chicago:
SmallWaters Corporation.
Banich, M. T. (2004). Cognitive neuroscience and neuropsychology. Boston: Houghton Mifflin.
Barnes, D. E., Yaffe, K., Satariano, W. A., & Tager, I. B. (2003). A
longitudinal study of cardiorespiratory fitness and cognitive function in
healthy older adults. Journal of American Geriatrics Society, 51, 459 –
465.
Bastin, C., & van der Linden, M. (2003). The contribution of recollection
and familiarity to recognition memory: A study of the effects of test
format and aging. Neuropsychology, 17, 14 –24.
Bennett, G. K., Seashore, H. G., & Wesman, A. G. (1997). Differential
Aptitude Test. San Antonio, TX: Psychological Corporation.
Bigler, E., Lowry, C. M., Kerr, B., Tate, D. F., Hessel, C. D., Earl, H. D.,
et al. (2003). Role of white matter lesions, cerebral atrophy, and APOE
on cognition in older persons with and without dementia: The Cache
County, Utah, study of memory and aging. Neuropsychology, 17, 339 –
352.
Burgess, P. W. (1997). Theory and methodology in executive function
research. In P. Rabbitt (Ed.), Methodology of frontal and executive
function (pp. 81–116). Hove, England: Psychology Press.
Carroll, J. B. (1993). Human cognitive abilities. Cambridge, England:
Cambridge University Press.
Collette, F., & van der Linden, M. (2002). Brain imaging of the central
executive component of working memory. Neuroscience and Biobehavioral Reviews, 26, 105–125.
de Jong, R. (2001). Adult age differences in goal activation and goal
maintenance. European Journal of Cognitive Psychology, 13, 71– 89.
Delis, D. C., Kaplan, E., & Kramer, J. H. (2001). Delis–Kaplan Executive
Function System. San Antonio, TX: Psychological Corporation.
Denchla, M. B. (1995). A theory and model of executive function: A
neuropsychological perspective. In G. R. Lyon & N. A. Krasnegor
(Eds.), Attention, memory, and executive function (pp. 263–278). Baltimore: Brooks.
Duncan, J. (1994). Attention, intelligence, and the frontal lobes. In M. S.
Gazzaniga (Ed.), The cognitive neurosciences (pp. 721–733). Cambridge, MA: MIT Press.
544
SALTHOUSE
Duncan, J., Emslie, H., Williams, P., Johnson, R., & Freer, C. (1996).
Intelligence and the frontal lobe: The organization of goal-directed
behavior. Cognitive Psychology, 30, 257–303.
Duncan, J., Johnson, R., Swales, M., & Freer, C. (1997). Frontal lobes after
head injury: Unity and diversity of function. Cognitive Neuropsychology, 14, 713–741.
Ekstrom, R. B., French, J. W., Harman, H. H., & Dermen, D. (1976).
Manual for kit of factor-referenced cognitive tests. Princeton, NJ: Educational Testing Service.
Elliott, R. (2003). Executive functions and their disorders. British Medical
Bulletin, 65, 49 –59.
Enders, C. K., & Bandalos, D. L. (2001). The relative performance of full
information maximum likelihood estimation for missing data in structural equation models. Structural Equation Modeling, 8, 430 – 457.
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-Mental
State: A practical method for grading the cognitive state of patients for
the clinician. Journal of Psychiatric Research, 12, 189 –198.
Foster, J. K., Black, S. E., Buck, B. H., & Bronskill, M. J. (1997). Ageing
and executive functions: A neuroimaging perspective. In P. Rabbitt
(Ed.), Methodology of frontal and executive function (pp. 117–134). East
Sussex, England: Psychology Press.
Friedman, N. P., & Miyake, A. (2004). The relations among inhibition and
interference control functions: A latent variable analysis. Journal of
Experimental Psychology: General, 133, 101–135.
Hambrick, D. Z., Salthouse, T. A., & Meinz, E. J. (1999). Predictors of
crossword puzzle proficiency and moderators of age– cognition relations. Journal of Experimental Psychology: General, 128, 131–164.
Heaton, R. K., Chelune, G. J., Talley, J. L., Kay, G. C., & Curtis, G. (1993).
Wisconsin Card Sorting Test manual: Revised and expanded. Odessa,
FL: Psychological Assessment Resources.
Lezak, M. (1995). Neuropsychological assessment (3rd ed.). New York:
Oxford University Press.
Malloy, P. F., Cohen, R. A., & Jenkins, M. A. (1998). Frontal lobe function
and dysfunction. In P. J. Snyder & P. D. Nussbaum (Eds.), Clinical
neuropsychology: A pocket handbook for assessment (pp. 573–590).
Washington, DC: American Psychological Association.
Meinz, E. J., & Salthouse, T. A. (1998). The effects of age and experience
on memory for visually presented music. Journals of Gerontology,
Series B: Psychological Sciences and Social Sciences, 53, P60 –P69.
Mittenberg, W., Seidenberg, M., O’Leary, D. S., & DiGiulio, D. V. (1989).
Changes in cerebral functioning associated with normal aging. Journal
of Clinical and Experimental Neuropsychology, 11, 918 –932.
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A.,
& Wager, T. D. (2000). The unity and diversity of executive functions
and their contributions to complex “frontal lobe” tasks: A latent variable
analysis. Cognitive Psychology, 41, 49 –100.
Mungas, D., Reed, B. R., & Kramer, J. H. (2003). Psychometrically
matched measures of global cognition, memory, and executive function
for assessment of cognitive decline in older persons. Neuropsychology, 17, 380 –392.
Perry, R. J., & Hodges, J. R. (1999). Attention and executive deficits in
Alzheimer’s disease: A critical review. Brain, 122, 383– 404.
Phillips, L. H. (1997). Do “frontal tests” measure executive function?
Issues of assessment and evidence from fluency tests. In P. Rabbitt (Ed.),
Methodology of frontal and executive function (pp. 191–213). Hove,
England: Psychology Press.
Rabbitt, P. (1997). Introduction: Methodologies and models in the study of
executive function. In P. Rabbitt (Ed.), Methodology of frontal and
executive function (pp. 1–38). Hove, England: Psychology Press.
Raven, J. (1962). Advanced Progressive Matrices: Set II. London: H. K.
Lewis.
Reitan, R. M., & Wolfson, D. (1994). A selective and critical review of
neuropsychological deficits and the frontal lobes. Neuropsychology Review, 4, 161–198.
Robbins, T. W., James, M., Owen, A. M., Sahakian, B. J., McInnes, L., &
Rabbitt, P. (1997). A neural systems approach to the cognitive psychol-
ogy of ageing using the CANTAB battery. In P. Rabbitt (Ed.), Methodology of frontal and executive function (pp. 215–238). Hove, England:
Psychology Press.
Royall, D. R., Cordes, J. A., & Polk, M. (1998). CLOX: An executive clock
drawing task. Journal of Neurology, Neurosurgery, and Psychiatry, 64,
588 –594.
Royall, D. R., Lauterbach, E. C., Cummings, J. L., Reeve, A., Rummans,
T. A., Kaufer, D. I., et al. (2002). Executive control function: A review
of its promises and challenges for clinical research. Journal of Neuropsychiatry and Clinical Neurosciences, 14, 377– 405.
Ruff, R. M. (1996). Ruff Figural Fluency Test. Odessa, FL: Psychological
Assessment Resources.
Salthouse, T. A. (1992). Why do adult age differences increase with task
complexity? Developmental Psychology, 28, 905–918.
Salthouse, T. A. (1993a). Influence of working memory on adult age
differences in matrix reasoning. British Journal of Psychology, 84,
171–199.
Salthouse, T. A. (1993b). Speed and knowledge as determinants of adult
age differences in verbal tasks. Journals of Gerontology, Series B:
Psychological Sciences and Social Sciences, 48, P29 –P36.
Salthouse, T. A. (1994a). Aging associations: Influence of speed on adult
age differences in associative learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20, 1486 –1503.
Salthouse, T. A. (1994b). The nature of the influence of speed on adult age
differences in cognition. Developmental Psychology, 30, 240 –259.
Salthouse, T. A. (1995). Influence of processing speed on adult age
differences in learning. Swiss Journal of Psychology, 54, 102–112.
Salthouse, T. A. (1996). General and specific speed mediation of adult age
differences in memory. Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 51, P30 –P42.
Salthouse, T. A. (2001a). Attempted decomposition of age-related influences on two tests of reasoning. Psychology and Aging, 16, 251–263.
Salthouse, T. A. (2001b). Structural models of the relations between age
and measures of cognitive functioning. Intelligence, 29, 93–115.
Salthouse, T. A. (2004). Localizing age-related individual differences in a
hierarchical structure. Intelligence, 32, 541–561.
Salthouse, T. A., Atkinson, T. M., & Berish, D. E. (2003). Executive
functioning as a potential mediator of age-related cognitive decline in
normal adults. Journal of Experimental Psychology: General, 132, 566 –
594.
Salthouse, T. A., & Babcock, R. L. (1991). Decomposing adult age
differences in working memory. Developmental Psychology, 27, 763–
776.
Salthouse, T. A., Babcock, R. L., Skovronek, E., Mitchell, D. R. D., &
Palmon, R. (1990). Age and experience effects in spatial visualization.
Developmental Psychology, 26, 128 –136.
Salthouse, T. A., Berish, D. E., & Siedlecki, K. (2004). Construct validity
and age sensitivity of prospective memory. Memory & Cognition, 32,
1133–1148.
Salthouse, T. A., & Ferrer-Caja, E. (2003). What needs to be explained to
account for age-related effects on multiple cognitive variables? Psychology and Aging, 18, 91–110.
Salthouse, T. A., Fristoe, N. M., Lineweaver, T. T., & Coon, V. E. (1995).
Aging of attention: Does the ability to divide decline? Memory &
Cognition, 23, 59 –71.
Salthouse, T. A., Fristoe, N., McGuthry, K., & Hambrick, D. Z. (1998).
Relation of task switching to age, speed, and fluid intelligence. Psychology and Aging, 13, 445– 461.
Salthouse, T. A., Fristoe, N., & Rhee, S. H. (1996). How localized are
age-related effects on neuropsychological measures? Neuropsychology, 10, 272–285.
Salthouse, T. A., Hambrick, D. Z., Lukas, K. E., & Dell, T. C. (1996).
Determinants of adult age differences in synthetic work performance.
Journal of Experimental Psychology: Applied, 2, 305–329.
EXECUTIVE FUNCTIONING
Salthouse, T. A., Hancock, H. E., Meinz, E. J., & Hambrick, D. Z. (1996).
Interrelations of age, visual acuity, and cognitive functioning. Journals
of Gerontology, Series B: Psychological Sciences and Social Sciences, 51, P317–P330.
Salthouse, T. A., McGuthry, K. E., & Hambrick, D. Z. (1999). A framework for analyzing and interpreting differential aging patterns: Application to three measures of implicit learning. Aging, Neuropsychology,
and Cognition, 6, 1–18.
Salthouse, T. A., & Meinz, E. J. (1995). Aging, inhibition, working
memory, and speed. Journals of Gerontology, Series B: Psychological
Sciences and Social Sciences, 50, P297–P306.
Salthouse, T. A., & Mitchell, D. R. D. (1990). Effects of age and naturally
occurring experience on spatial visualization performance. Developmental Psychology, 26, 845– 854.
Salthouse, T. A., Nesselroade, J. R., & Berish, D. E. (2004). Are single
assessments sufficient? The case for measurement burst designs. Unpublished manuscript.
Salthouse, T. A., & Prill, K. A. (1987). Inferences about age impairments
in inferential reasoning. Psychology and Aging, 2, 43–51.
Salthouse, T. A., Toth, J., Daniels, K., Parks, C., Pak, R., Wolbrette, M., &
Hocking, K. (2000). Effects of aging on the efficiency of task switching
in a variant of the Trail Making Test. Neuropsychology, 14, 102–111.
Salthouse, T. A., Toth, J. P., Hancock, H. E., & Woodard, J. L. (1997).
Controlled and automatic forms of memory and attention: Process purity
and the uniqueness of age-related influences. Journals of Gerontology,
Series B: Psychological Sciences and Social Sciences, 52, P216 –P228.
Siedlecki, K. L., Salthouse, T. A., & Berish, D. E. (2005). Is there anything
special about the aging of source memory? Psychology and Aging, 20,
19 –32.
545
Stern, R. A., & White, T. (2003). Neuropsychological Assessment Battery
administration, scoring, and interpretation manual. Lutz, FL: Psychological Assessment Resources.
Tranel, D., Anderson, S. W., & Benton, A. (1994). Development of the
concept of “executive function” and its relationship to the frontal lobes.
In F. Boller & J. Grafman (Eds.), Handbook of neuropsychology (Vol. 8,
pp. 125–148). Amsterdam: Elsevier.
Troyer, A. K., Graves, R. E., & Cullum, K. M. (1994). Executive functioning as a mediator of the relationship between age and episodic
memory in healthy aging. Aging and Cognition, 1, 45–53.
Tuokko, H., & Hadjistavropoulos, T. (1998). An assessment guide to
geriatric neuropsychology. Mahwah, NJ: Erlbaum.
Verghese, J., Lipton, R. B., Katz, M. J., Hall, C. B., Derby, C. A.,
Kuslansky, G., et al. (2003). Leisure activities and the risk of dementia
in the elderly. The New England Journal of Medicine, 348, 2508 –2516.
Wechsler, D. (1997a). Wechsler Adult Intelligence Scale—Third Edition.
San Antonio, TX: Psychological Corporation.
Wechsler, D. (1997b). Wechsler Memory Scale—Third Edition. San Antonio, TX: Psychological Corporation.
Woodcock, R. W., & Johnson, M. B. (1990). Woodcock–Johnson PsychoEducational Battery—Revised. Allen, TX: DLM.
Zachary, R. A. (1986). Shipley Institute of Living Scale—Revised. Los
Angeles: Western Psychological Services.
Received January 12, 2004
Revision received September 23, 2004
Accepted October 25, 2004 䡲
Download
Related flashcards

Statistical theory

24 cards

Statistical theory

25 cards

Statistical inference

13 cards

Statistical inference

12 cards

British statisticians

81 cards

Create Flashcards